Destination film from ten location photos with Wan 3.0 on Sume

Pick ten photos of the place, send them as input_references to wan-3.0, and ask for one 30-second film with a beginning, a turn and an end. On Sume that job costs $3.75 at 720p or $7.50 at 1080p. Alibaba lists tourism and destination marketing as a Wan 3.0 use case and says the model reproduces real scenes. That is the vendor's claim, so the plan below treats every frame as generated and checks it against the photos.

Choose the ten photos by role

• Three wide landscapes at different times of day.
• Three landmarks or buildings, each photographed straight on.
• Two food or market images with people in them.
• Two detail shots, such as a doorway, a boat or a sign.
• Keep one orientation across the set so the film does not need to choose between 16:9 and 9:16.

Cost of the options

Wan 3.0 lists $0.05, $0.10 and $0.20 per second at the three resolutions, and Sume bills the list price times 1.25. Thirty seconds is 30 x $0.125 = $3.75 at 720p.

Write the prompt as an arc

Morphic's page makes a useful point about 30-second clips: length without a turn is just a slow clip, so the brief should say how the frame changes. For a town, that can be harbour at dawn, market at midday, hillside lit at dusk. Morphic also advises labelling each reference in the prompt. Sume's docs describe references as visual guidance and document no tag syntax, so describe each image in words, such as the blue-shuttered harbour house from the first photo.

Sume's catalog entry for wan-3.0 accepts up to 10 reference images, 5 reference videos and 5 reference audios per job, and a job can run 2 to 30 seconds. Aspect ratios include 16:9, 9:16, 1:1, 4:3 and 3:4. A vertical cut for social is a second job, not a crop.

A shot list that fits thirty seconds

Thirty seconds holds roughly five beats of six seconds each: arrival, the main landmark, the street or market, a food or people moment, and a closing wide shot. Write them into the prompt in order with a transition phrase between each, such as the light warms as the camera climbs the hill. The model decides the cuts, so a prompt with five clear beats gives it something to follow, while a prompt that lists ten attractions gives it nothing to prioritise.

If one beat matters more than the rest, give it its own job. A hero shot of the landmark at 1080p for 6 s costs $1.50, and you can join it to the rest of the film in an editor.

Checks before publishing

• Compare every landmark with its photo. References steer the look but are not exact frames.
• Do not rely on signs or menus in frame. The vendor lists on-screen text accuracy as a known limitation, so add real titles in an editor instead.
• If you need an exact first shot, send that photo as frame_images with frame_type first_frame. Sume then treats the job as image-to-video and ignores references as references, so you must choose.
• A 480p draft at $1.88 is worth one run before the 1080p final.